Aggrey Lelei
Head of Tech
I'm a Tech Lead specializing in Spring Boot enterprise systems and distributed architecture. I Built and maintain a cloud-native real-time chat, voice call + notification workflow, handling ~400 concurrent peak users with event-based coordination, backpressure-aware processing, and production monitoring.
About Aggrey
Currently the Head of Technology at Tiret Resorts . I lead a team of five engineers and own the technical direction, architecture, and delivery of the cloud-native communication and logistics platform used by NGOs to coordinate regional operations.
Under my management the platform increased NGO productivity and regional engagement by ~55%, supporting ~400 concurrent users with sub-150 ms response times under peak load.
The system delivers real-time chat, voice calls, secure notifications, and intelligent summarization, built on event-driven coordination, backpressure-aware processing, and production-grade monitoring. I design and operate Spring Boot–based distributed services with clear boundaries and strong observability.
I also integrate Embabel to enable agentic and retrieval-augmented workflows grounded in live system data, while leading hiring, mentoring, and architectural decisions to ensure the platform scales reliably.
Projects
Exploring distributed systems alongside Embabel in this AI era led me to the realization that AI orchestration is rising as LLM performance gains plateau. Inspired by Embabel, a more disciplined evolution of LangChain and Moltbot, I’ve been building Sparrow-X, which prioritizes instructional rigor and architectural clarity over prompt-driven improvisation.
The Project is still under construction
Agentic & LLM Execution: Designed and operated an Embabel-based agentic execution layer orchestrating retrieval, planning, reasoning, and tool invocation across distributed services, with measured latency budgets, deterministic execution graphs, and observability hooks to evaluate goal-driven LLM workflows under production constraints.
• Scale & Performance: Should sustain ~5,000 rps with p95 sub-200 ms latency across read/write-heavy and fan-out workloads.
• Architecture: Production microservices platform using Docker, API Gateway, service-to-service mTLS (Linkerd), and an event-driven bus (Kafka) with backpressure and retries.
• Reliability & Resilience: Built-in fault isolation using Resilience4j (timeouts, retries, circuit breakers, rate limiting) to maintain SLOs under partial failures.
• Observability & Operations: Full-stack observability with Prometheus, Tempo, and Loki, enabling latency decomposition, saturation analysis, and fast incident triage.
Tech Stack:
Technical Writing
I write technical articles and guides for the developer community, spanning applied topics like distributed systems and theoretical foundations such as information theory.
How Increasing Workload Complexity, Not Smarter Tools, Revealed the Irreplaceable Role of Humans.
Why bow thy head beneath the name of fraud? Had Claude Elwood Shannon risen in this AI century, where thoughts fly swifter than breath, information & entropy reckon faster than the pulse, he too would stagger beneath excess of knowing; lost among glowing screens and infinite paths. He would doubt, as thou dost now, amid networks without end and truths summoned at a keystroke. Yet from such doubt are new dominions seized; for uncertainty is but the shadow cast by a mind straining to outrun its tools and shape the future they obey.
You owe thyself the luxury of locking in.
- Probably your super ego
© 2026 Aggrey Lelei.